Towards fully automated remote ecosystem monitoring

Natural ecosystems around the world are being impacted by human activity at an ever-increasing rate. However, we still don’t fully understand the true extent of our actions on these complex systems, limiting our ability to develop sustainable, well informed best practices.

Much of the problem is in collecting sufficient amounts of data from environments which are often difficult for scientists to access and survey thoroughly (e.g. polar regions, tropical rainforests, savannas). Therefore, we have been working on methods of fully-automating ecosystem monitoring in a cost-effective way, that will provide huge amounts of data on the health of a remote ecosystem over long time periods, with minimal effort required by field scientists to maintain the system.

Our first step towards this goal has been to develop a device that continuously records data from a variety of sensors (microphones, cameras, humidity sensors etc.) and uploads the data to the internet directly from the field using a standard mobile phone internet connection. The device is also powered by a solar panel setup, meaning that battery replacements are unnecessary. In theory, once initially set up, this device can sit out in a remote field site indefinitely, with the data sent straight to scientists almost instantaneously. Mobile phone connections are patchy at best in remote locations so a key challenge was to have a low-power system that could opportunistically exploit the available mobile signal.

The kit is cheap and open source so you can make your own and you're welcomed to have an explore: http://www.rpi-eco-monitoring.com and, if you subscribe, you can read more in a recent New Scientist article.  You can read our full paper for free in Methods in Ecology and Evolution for more details - https://besjournals.onlinelibrary.wiley.com/doi/abs/10.1111/2041-210X.13089 

Of particular interest to our group has been using audio to identify calling animals in the tropical forests of Borneo. We work at the SAFE project site in Sabah where ecologists from around the world investigate the effect of logging and the oil palm industry on the biodiversity of these ancient rainforests. A growing network of 12 monitoring devices are currently scattered around the SAFE landscape (see Rob Ewers's work) in areas varying from old-growth forest (with almost no impact from humans) to oil palm plantations. 

A real-time acoustic monitoring unit, deployed in the tropical forests of Sabah, Borneo at the SAFE project site. Data used from these devices helps investigate the effect of the oil palm industry on the biodiversity found in the region.

We are developing algorithms using a wide array of machine learning techniques that will automatically listen to the masses of audio from these monitoring devices in Borneo and give us a real-time measure of the biodiversity in the different forest locations. With a finer scale understanding of the full human impact on these fragile ecosystems we can help inform better sustainable practices for the oil-palm industry to minimise their damage on the threatened species of this region. Sarab, Lorenzo, Rob and Nick

 

Aging in the Variance: increasing variation between egg-cells with age

DNA in mitochondria, the powerhouses of the cell, is passed down from mother to child. But there are many mitochondria in each cell, and these mitochondria may have different genetic features. If a mother carries a mixture of mitochondrial DNA (mtDNA) types, this can make it hard to say which features their children will inherit. For mothers carrying a disease-causing mtDNA mutation, this makes family planning and clinical therapies challenging.

In particular, the role of a mother's age has long been a mystery. Is the probability of a child inheriting a particular mtDNA feature higher when mothers are younger or older? An answer to this question could help plan clinical strategies to improve fertility and prevent the inheritance of deadly mitochondrial disease.

Joerg Burgstaller and colleagues (see our blog entry here) previously made a type of mouse that contains two (apparently benignly differing) types of mitochondria (and their genetic material mtDNA) in every cell. This means that their egg-cells that make the next generation also have both types of mtDNA present. Our latest paper investigates how the proportion of the two types of mtDNA varies within the ovary, finding that cells become more and more variable with time.

To address this, we worked with our excellent collaborators with a combination of maths, statistics, and experiment. Our collaborators used cutting-edge technology to reveal the proportions of two-types of mtDNA in the egg cells of mother mice at a wide range of ages, and in the litters of offspring the mothers produced. This experimental work was the largest-scale study of mammalian mtDNA that we're aware of, involving thousands of observations throughout lifetimes and between generations. In concert, we developed a mathematical model describing the changes to, and inheritance of, mtDNA from mother to offspring. We combined the model and data to learn how different biological processes affect mtDNA through and between generations.

We found that the cell-to-cell variability in the proportions of the two types of mtDNA dramatically increased as mothers aged. This means that the probability of inheriting more extreme -- both lower and higher -- levels of a genetic feature increases for older mothers. We also found that different mtDNA mixtures were inherited in different ways - with some mtDNA types favoured for inheritance and some disfavoured. We used our findings to create a way to predict how the risk that offspring would inherit disease-causing mtDNA features changes over time. Moving forward, we're aiming to harness these powerful ways of using large datasets to describe and predict the dynamics of mtDNA inheritance in humans, and to learn what it is about these mtDNA types that predicts their evolution across generations. You can read the article “Large-scale genetic analysis reveals mammalian mtDNA heteroplasmy dynamics and variance increase through lifetimes and generations” for free in Nature Communications here. Iain, Joerg and Nick

How cells adapt to progressive increase in mitochondrial mutation

Mitochondria produce the cell's major energy currency: ATP. If mitochondria become dysfunctional, this can be associated with a variety of devastating diseases, from Parkinson's disease to cancer. Technological advances have allowed us to generate huge volumes of data about these diseases. However, it can be a challenge to turn these large, complicated, datasets into basic understanding of how these diseases work, so that we can come up with rational treatments.

We were interested in a dataset (see here) which measured what happened to cells as their mitochondria became progressively more dysfunctional. A typical cell has roughly 1000 copies of mitochondrial DNA (mtDNA), which contains information on how to build some of the most important parts of the machinery responsible for making ATP in your cells. When mitochondrial DNA becomes mutated, these instructions accumulate errors, preventing the cell's energy machinery from working properly. Since your cells each contain about 1000 copies of mitochondrial DNA, it is interesting to think about what happens to a cell as the fraction of mutated mitochondrial DNA (called 'heteroplasmy') gradually increases.  We used maths to try and explain how a cell attempts to cope with increasing levels of heteroplasmy, resulting in a wealth of hypotheses which we hope to explore experimentally in the future.

The central idea arising from our analysis of this large dataset is that cells seem to attempt to maintain the number of normal mtDNAs per cell volume as heteroplasmy initially increases from 0% mutant. We suggest they do this by shrinking their size. By getting smaller, cells are able to reduce their energy demands as the fraction of mutant mtDNA increases, allowing them to balance their energy budget and maintain energy supply = demand. However, cells can only get so small and eventually the cell must change its strategy. At a critical fraction of mutated mtDNA (h* in the cartoon above), we suggest that cells switch on an alternative energy production mode called glycolysis. This causes energy supply to increase, and as a result, cells grow larger in size again. These ideas, as well as experimental proposals to test them, are freely available in the Biochemical Journal "Mitochondrial DNA Density Homeostasis Accounts for a Threshold Effect in a Cybrid Model of a Human Mitochondrial Disease". Juvid, Iain and Nick

What we learn from the learning rate

Cells need to sense their environment in order to survive. For example, some cells measure the concentration of food or the presence of signalling molecules. We are interested in studying the physical limits to sensing with limited resources, to understand the challenges faced by cells and to design synthetic sensors.

We have recently published a paper 'What we learn from the learning rate' (free version) where we explore the interpretation of a metric called 'the learning rate' that has been used to measure the quality of a sensor (e.g here). Our motivation is that in this field a number of metrics (a metric is a number you can calculate from the properties of the sensor that, ideally, tells you how good the sensor is) have been applied to make some statement about the quality of sensing, or limits to sensory performance. For example, a limit of particular interest is the energy required for sensing. However, it is not always clear how to interpret these metrics. We want to find out what the learning rate means. If one sensor has a higher learning rate than another what does that tell you? 

The learning rate is defined as the rate at which changes in the sensor increase the information the sensor has about the signal. The information the sensor has about the signal is how much your uncertainty about the state of the signal is reduced by knowing the state of the sensor (this is known as the mutual information). From this definition, it seems plausible that the learning rate could be a measure of sensing quality, but it is not clear. Our approach is a test to destruction – challenge the learning rate in a variety of circumstances, and try to understand how it behaves and why.

To do this we need a framework to model a general signal and sensor system. The signal hops between discrete states and the sensor also hops between discrete states in a way that follows the signal. A simple example is a cell using a surface receptor to detect the concentration of a molecule in its environment.

The figure shows such a system. The circles represent the states and the arrows represent transitions between the states. The signal is the concentration of a molecule in the cell’s environment. It can be in two states; high or low, where high is double the concentration of low. The sensor is a single cell surface receptor, which can be either unbound or bound to a molecule. Therefore, the joint system can be in four different states. The concentration jumps between its states with rates that don’t depend on the state of the sensor. The receptor becomes unbound at a constant rate and is bound at a rate proportional to the molecule concentration. 

We calculated the learning rate for several systems, including the one above, and compared it to the mutual information between the signal and the sensor (the mutual information is a refined measure of correlation). We found that in the simplest case, shown in the figure, the learning rate essentially reports the correlation between the sensor and the signal and so it is showing you the same thing as the mutual information. In more complicated systems the learning rate and mutual information show qualitatively different behaviour. This is because the learning rate actually reflects the rate at which the sensor must change in response to the signal, which is not, in general, the equivalent to the strength of correlations between the signal and sensor. Therefore, we do not think that the learning rate is useful as a general metric for the quality of a sensor. Rory, Nick and Tom